基於Kinect的手勢追踪來開發Wayang Kulit劇院;KINECT-BASED HAND GESTURE TRACKING FOR DEVELOPING WAYANG KULIT THEATER

NCUIR > College of Electrical Engineering & Computer Science > Graduate Institute of Computer Science and Information Engineering > Electronic Thesis & Dissertation > Item 987654321/74696

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/74696

Title:	基於Kinect的手勢追踪來開發Wayang Kulit劇院;KINECT-BASED HAND GESTURE TRACKING FOR DEVELOPING WAYANG KULIT THEATER
Authors:	歐文尼斯;Yohannes, Ervin
Contributors:	資訊工程學系
Keywords:	皮影戲;模板匹配;手勢辯識;深度傳感器;Wayang Kulit;template matching;hand gesture recognition;depth sensor
Date:	2017-07-27
Issue Date:	2017-10-27 14:36:32 (UTC+8)
Publisher:	國立中央大學
Abstract:	在印度尼西亞有許多傳統的遊戲和劇院，這些遊戲跟劇院適合所有年齡層的人們，最有名就是皮影戲，它是用戲偶的影子動作來進行表演，能夠表演不同類型的劇本，像是喜劇、歷史、愛情等等。其中在印度尼西亞受大家歡迎的就是愛情劇，叫做Ramayana的故事，在這個故事中有五個角色，包含Anoman, Sinta, Rama, Laksma, 和Rahwana。Rahwana是屬於反派的一方，其他人是正派的一方。我們的皮影戲利用手勢辯來操控建構出來的虛擬角色，引入叫做blender的工具來創造2D或3D的物件，選擇在Unity的環境中來進行開發，Unity能夠結合Kinect的相關模組進行更多有趣的變化。首先，因為需要兩隻手來操控戲偶，需要將手勢進行左右手的分群來進行辯識，主要使用的方法為DBSCAN分群法，能夠自動且有效率的分群，在DBSCAN的方法中Eps和MinPts是對於能否得到好的結果重要的兩個參數。右手和左手的群集會先搜尋各自的平均數，這個過程也可以叫做標記，因為系統會選擇哪一個為平均數1和平均數2。找完平均數後，就能開始追蹤手部動作，平均數1和平均數2之間的距離我們設置了三種情況能夠觸發不同效果，第一種距離就是當其他皮影戲偶遇到Rahwana時，如果距離在50~100之間就會觸發打鬥，第二種距離就是當距離為100~200之間，會撥放動畫在每個皮影戲偶上，最後一種就是當距離超過200時，就讓每個皮影戲偶正常表演。我們利用模板匹配的方法，設定不同手部方向的模板來進行手勢辯識，每個方向都能夠對皮影戲偶產生不同的效果。本研究的貢獻為評估DBSCAN門檻值、手勢辯識和皮影戲表演的使用者體驗。DBSCAN門檻值在25時對於我們系統手勢辯識的部分是最好的，使用者體驗評估的部分包含了6項東西，吸引力、效能表現、簡明性、可靠性、刺激性和新奇程度，我們觀察了20個人，他們對於皮影戲有不同等級的熟悉程度，根據觀察後的結果，以上因素都表現良好。我們的系統未來能夠改善的東西包含了: 使用其他的辯識方法、變更故事內容、改變皮影戲偶的表演方式等等。;In Indonesia, there are many traditional games and theater. Both games and theater can be playing by all person from a kid until mature. The popular theater is playing doll and shadow together by person and It is called Wayang Kulit theater. Many storytellings in Wayang Kulit from comedy, history, romance and other. Romance story is popular story than another story which is Ramayana story. In this story, there is five character including Anoman, Sinta, Rama, Laksma, and Rahwana. In the Wayang Kulit theater usually there are Gunungan to give sign background or storytelling changed. Rahwana is a bad character and others are a good character. The Wayang Kulit theater using hand gesture recognition for playing. The 3D object made from Blender which is blender is a tool for designing 2D or 3D object, many features inside for designing or modeling, give the texture of object and etc. Blender is free than another tool for the same purpose and it is user-friendly so user can be understanding the program or tool quickly. The programming using C# by related Unity. Unity is game development tool and many great games made from Unity because of so many features, the tool so interactive and the important thing it can be relating to Kinect. Firstly, in theater need two hands to generate or move Wayang Kulit so need clustering method for cluster right and left hand. The clustering method using DBSCAN clustering because can be cluster automatically and efficiently than another method. There are 2 important variables in DBSCAN including Eps and MinPts which are need validate threshold to get good result of the cluster. The right and left hand will search about each median for next processing. The search median can be namely labeling processing because the system will be choosing which one median1 and median2. After finding each median the processing can recognize about gesture movement. The gesture is choosing the Wayang Kulit in above there are six objects. After that, the distance between median1 and median2 will be compared. There is three distance in this performance. The first distance is fighting effect which is distance range is 50 to 100 that occur if Wayang Kulit meets the Rahwana so show the fighting ability. The second distance is playing the animation in each Wayang Kulit and the distance is 100 to 200. The last is important gesture because each Wayang Kulit can show each ability and the distance range more than 200. The last, gesture recognition using template matching method which is there are three templates including down, right, left direction. Each direction will be shown different ability in each Wayang Kulit but only two Wayang Kulit has three abilities and the other just has one ability. The evaluation of this research consists of DBSCAN threshold, gesture recognition, and UI / UX evaluation. The DBSCAN threshold evaluation looks for the best threshold for clustering right and left hand. From the result, the best threshold is 25 because it can be clustered median1 and median2 exactly. For the gesture recognition evaluation, we will observe 20 people from various knowledge discipline for playing Wayang Kulit performance and the gesture recognition focus in the last threshold that distance range more than 200. The average of all result above 90% and the best result is median 2 on gesture 1 which is result 100%. The last UI / UX evaluation which is 20 people of hand gesture recognition will get a questionnaire and ask to fill it after playing Wayang Kulit performance. The UI / UX must contain 6 factors including attractiveness, efficiency, perspicuity, dependability, stimulation, and novelty. From the evaluation result overall is good for our system. The only improvement in the future the result is very good it means our system can be interested people to developing our system in the future through using another method, change of story, change the Wayang Kulit performance, and others.
Appears in Collections:	[Graduate Institute of Computer Science and Information Engineering] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	294	View/Open

社群 sharing

Loading...